Is Rotakin credible in the digital age?
The Rotakin test was established over two decades ago, and was formulated in a time of analogue video. In today’s digital age, the target-based test presents challenges for installers and integrators. Benchmark considers the best approach, assisted by Steven Ryan of Axis Communications.
The Rotakin test target was designed approximately a quarter of a century ago to evaluate the performance of CCTV systems. It’s goal was to ensure that cameras were theoretically capable of producing suitable images of intruders. The test was used to assess whether or not individuals captured by video could be identified, and to determine how well users would be able to see camouflaged individuals in differing weather or lighting conditions.
The original Rotakin guidelines were used to establish a required percentage of the screen that a target should fill. Differing values were specified dependent upon the intended use of the video surveillance system.
These were 120 per cent for identification, delivering picture quality and detail sufficient to enable the identity of a subject to be established beyond reasonable doubt; not less than 50 per cent for recognition, giving a high degree of certainty whether or not the individual is someone seen previously; not less than 10 per cent for detection, giving a high degree of certainty a person is visible within an image, and not less than 5 per cent for monitoring, allowing a user to determine the number, direction and speed of movement of viewed individuals.
Elements of the original tests specified resolution in TV lines, which helped establish the required percentages as mentioned. These measurements established the dimensions of critical details.
The tests – based on a traditional analogue 4CIF image – assumed that all cameras had the same maximum resolution and displayed a full frame or field image. At the time, this was pretty much correct. However, the significantly wider and greater range of IP cameras with varying resolutions and advanced streaming options simply do not fit with such an assumption.
A step forwards?
With the advent of higher resolution IP cameras, in some circumstances it could be difficult to configure and commission a network camera in line with the Rotakin test requirements. With increasing IP surveillance cameras sales, EN 50132-7 was modified in June 2013, taking into consideration ‘megapixels per metre’.
This move was made in order to allow a direct comparison with the levels of detail in the prescribed percentages used for identification, recognition, detection and general monitoring.
Higher resolutions for HD cameras and multi-megapixel devices simply do not require the prescribed percentages, as elements of the scene can be digitally zoomed or identified as regions of interest, and despite this magnification will still deliver higher resolution images with more detail than a standard analogue camera ever will!
The change allowed installers, integrators and specifiers to determine what level of detail an IP camera would capture at specified distances, in line with the site operational requirements.
In principle, this approach is an accurate way of benchmarking the performance of an IP surveillance system in line with operational requirements. With the inclusion of the level of new detail in the standards documents, it meant that there was a comparison between the two technologies.
Of course, the standard and the addendums did not account for the flexibility of modern networked cameras, and in some cases there were issues relating to how the video was to be used. There were a number of difficulties adopting the new IP standards related to megapixels per metre in specific surveillance environments.
Whilst the new standards guaranteed that the level of detail captured was 100 per cent accurate, in some cases it was only beneficial for operational systems which were used for evidential purposes or when users retrospectively interrogated the viewed scene. Whilst there was the capacity to view the footage at appropriate detail levels after an event, this wasn’t always the case with live viewing.
When a surveillance system is being proactively operated, the level of detail obtained in a live view on a digital screen could be too small for an operator to distinguish objects accurately. The level of detail would, however, have been guaranteed when using a Rotakin test to commission the surveillance system. It is a case of the test process not being flexible enough to accommodate the advanced possibilities available from modern systems.
This has recently been identified and as a result an addendum has been made to the standards related to the testing and commissioning of IP surveillance systems. The result was that an appendix was added, stating that live view IP surveillance camera images must comply with the original Rotakin test and be shown as a percentage of screen height once again. This has obviously created some challenges to manufacturers and system integrators when trying to adhere to a standard that uses terminology in TV lines adopted for a fixed resolution analogue camera.
It may ne necessary to ensure a captured image fills a specific screen height so the level of detail can been seen when live, rather than as evidence at a later date. Installers, integrators and specifiers therefore may need to make an IP surveillance system work within a standard-based around an analogue technology.
A different approach
Today, many security installations use video analytics to automatically detect objects and people within a camera’s field of view. These automatic detections can then be sent to an operator’s screen for visual verification of the alarm event.
Due to the improvement in IP camera technology and the increase in pixel density now available, automatic detection systems are able to see much further, and to do so in more challenging conditions than ever before. This improved performance can, however, have a negative impact on the visual verification requirement of the operator.
In longer detection zones, some objects may appear too small on the operator’s screen, meaning that the secondary visual verification required by the operator is not always possible.
In a video image, when the target subject is near the camera they will be clearly noticeable by the operator as they fill a good percentage of the vertical screen height.
As shown in the right hand side image on page 53, if the subject is at the far end of the scene, even though an automatic motion detection system or video analytics engine may detect them, they fill a much smaller percentage of the vertical screen height. This may cause the operator some difficulties in providing a quick and accurate visual verification of the alarm event.
Because of this, many security surveillance system designs will require that when an automatic detection occurs, the object still fills a minimum proportion of the operator’s screen, thus enabling accurate secondary visual verification of the event. Of course, this requirement can introduce a degree of compromise.
Often the solution is to use a higher number of standard resolution cameras and shorten the detection zones. This ensures the subject fills a high percentage of the vertical screen height, enabling the required visual verification by the operator. Unfortunately, the customers who accept such an approach miss out on the potential cost saving benefits that come with using higher mega pixel cameras and longer detection zones. Also, for the installer, integrator and specifier, it can result in a system which is not financially viable!
One feature which many modern IP-based cameras allow is the selection of specific areas in an image, which can then be streamed as ‘virtual cameras’. Often referred to as view areas, areas of interest or a host of other names, these can be selected as an independent stream from a multi-streaming device.
Increasingly, modern VMS solutions are able to handle multiple streams from a single device, including ‘virtual cameras’, and often these can have very different configurations to other streams from a single camera. In effect, they can have all the typical configurations of a physical device – IVA, motion detection, alarm handling, event-actioned triggers, rules and actions – applied to them. Also, as they are independent from the other streams delivered by the camera, a higher degree of flexibility is on offer. This means that installers and integrators can decide which streams are displayed in response to events.
Subsequently, the use of ‘virtual cameras’ can allow the system to retain a single multi-megapixel camera, and can enjoy the benefits of IVA or motion detection, whilst still meeting requirements for live displays to show certain sized images of targets.
By using the multi-streaming functionality in combination with different automatic detection zones within a camera’s field of view, an event can trigger the display of a specific cropped area of the scene, enabling quicker and more accurate visual verification.
The original image is still streamed and recorded, if that is what the operational requirement dictates. However, the camera can also stream a ‘virtual camera’ created by cropping the area of the image in which far detection zones are located. The result is that any target which triggers an event will now appear as a much higher percentage of the vertical screen height. This ensures that an operator can clearly see any intruder and can act upon the event having verified it. Dependent upon camera specifications, different views can be created to address the needs of the system, allowing identification, recognition, detection and monitoring in different areas of the scene, if necessary.
These additional camera image streams require no additional hardware at the camera end (if a suitable device is selected). Depending on the VMS being used, typically no additional software licences are required either. This is because the multiple streams are generated from a single IP address (one camera).
With these extra ‘virtual camera’ views and multiple zones on the automatic detection system, the VMS can be configured to display different cropped streams depending on which zone is activated. This allows the benefits of advanced IP cameras to be realised, whilst also maximising the operators view, which ensures higher efficiency.
The Rotakin test is very much from a different age, but it still is relevant based upon the operator’s ability to react to live views. However, modern IP cameras do deliver the ability to exploit advanced technology and still meet the standards for displayed live images. Careful design can ensure best practice and higher performance!